Fast SVM training algorithm with decomposition on very large data sets

被引:141
作者
Dong, JX [1 ]
Krzyzak, A [1 ]
Suen, CY [1 ]
机构
[1] Concordia Univ, Dept Comp Sci & Software Engn, Ctr Pattern Recognit & Machine Intelligence, Montreal, PQ H3G 1M8, Canada
基金
加拿大自然科学与工程研究理事会;
关键词
support vector machines (SVMs); algorithm design and analysis; algorithm efficiency; machine learning; handwritten character recognition;
D O I
10.1109/TPAMI.2005.77
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
Training a support vector machine on a data set of huge size with thousands of classes is a challenging problem. This paper proposes an efficient algorithm to solve this problem. The key idea is to introduce a parallel optimization step to quickly remove most of the nonsupport vectors, where block diagonal matrices are used to approximate the original kernel matrix so that the original problem can be split into hundreds of subproblems which can be solved more efficiently. In addition, some effective strategies such as kernel caching and efficient computation of kernel matrix are integrated to speed up the training process. Our analysis of the proposed algorithm shows that its time complexity grows linearly with the number of classes and size of the data set. In the experiments, many appealing properties of the proposed algorithm have been investigated and the results show that the proposed algorithm has a much better scaling capability than Libsvm, SVMlight, and SVMTorch. Moreover, the good generalization performances on several large databases have also been achieved.
引用
收藏
页码:603 / 618
页数:16
相关论文
共 42 条
[11]   SVMTorch: Support vector machines for large-scale regression problems [J].
Collobert, R ;
Bengio, S .
JOURNAL OF MACHINE LEARNING RESEARCH, 2001, 1 (02) :143-160
[12]  
Cristianini N., 2000, Intelligent Data Analysis: An Introduction, DOI 10.1017/CBO9780511801389
[13]   Training invariant support vector machines [J].
Decoste, D ;
Schölkopf, B .
MACHINE LEARNING, 2002, 46 (1-3) :161-190
[14]   MAXIMUM LIKELIHOOD FROM INCOMPLETE DATA VIA EM ALGORITHM [J].
DEMPSTER, AP ;
LAIRD, NM ;
RUBIN, DB .
JOURNAL OF THE ROYAL STATISTICAL SOCIETY SERIES B-METHODOLOGICAL, 1977, 39 (01) :1-38
[15]   A fast SVM training algorithm [J].
Dong, JX ;
Suen, CY .
INTERNATIONAL JOURNAL OF PATTERN RECOGNITION AND ARTIFICIAL INTELLIGENCE, 2003, 17 (03) :367-384
[16]  
DONG JX, 2003, P INT WORKSH ART NEU
[17]  
DONGARRA JJ, 1990, ACM T MATH SOFTWARE, V16, P1, DOI 10.1145/77626.79170
[18]   Efficient SVM regression training with SMO [J].
Flake, GW ;
Lawrence, S .
MACHINE LEARNING, 2002, 46 (1-3) :271-290
[19]   CALCULATION OF BAYES RECOGNITION ERROR FOR 2 MULTIVARIATE GAUSSIAN DISTRIBUTIONS [J].
FUKUNAGA, K ;
KRILE, TF .
IEEE TRANSACTIONS ON COMPUTERS, 1969, C 18 (03) :220-&
[20]  
Fukunaga K., 1990, INTRO STAT PATTERN R